Don’t worry — the changes were not enough to be detected by the human ear. But a recognition of these shifts could prove to be very useful. The study was the work of a digital health startup named Klick Labs, which is developing voice-recognition software that may be able to diagnose type 2 diabetes simply by listening to adults speak a few sentences.
Klick scientists took the recordings of several hundred adults, both with and without diabetes, and fed them to an artificial intelligence (AI) model capable of analyzing each snippet. By examining the acoustic features of each voice, the AI system learned that factors such as vocal pitch (in women), strength (in men), and variability (in both) seem to differ meaningfully in people with diabetes. This allows the program to make a surprisingly accurate guess as to whether someone has diabetes — and it only needs to hear you talk for 10 seconds.
The results: The AI model “has 89 percent accuracy for women and 86 percent for men,” according to a press release. The designers hope that their innovation could become a useful tool for the diagnosis of type 2 diabetes, which currently relies on blood tests that are more invasive and time-consuming. Voice recognition could potentially be a faster and less expensive screening alternative, one that could be done even without an office visit.
“Current methods of detection can require a lot of time, travel, and cost. Voice technology has the potential to remove these barriers entirely,” states Jaycee Kaufman, a research scientist at Klick Labs and the primary author of the new paper.
Diabetes, Nerve Damage, and the Throat
Actually, this isn’t the first research to consider the effect that diabetes can have on the voice.
A 2012 study found that people with diabetes have a higher degree of hoarseness — though only those “with poor glycemic control and with neuropathy.”
A 2019 systematic review found that 12.5 percent of people with diabetes (1 in 8) have “voice problems,” far higher than in the general population, ranging from hoarseness and straining to excessive throat clearing, annoying coughing, and the sensation of a lump in one’s throat.
Such issues appear to be more common in people with neuropathy (nerve damage). A 2022 review explains that chronically elevated blood glucose levels cause nerve dysfunction throughout the entire body — including the throat and neck. The nerve fibers of people with diabetes experience “progressive destruction,” with consequences such as weakness, reduced sensation, and ataxia (loss of coordination). This dysfunction, presumably, can alter your voice.
More Experimentation Needed
Klick Labs’ innovation is not ready for prime time yet. Medical authorities will need to see a much more robust proof of accuracy before they can consider recommending the voice recognition program. Yan Fossat, vice president of Klick Labs, says that follow-up validation studies will require “individuals with different characteristics than our original study, with different demographics, from different regions of the world.”
It also seems reasonable to guess that, like other complications, voice changes are more prevalent and substantial in people with diabetes of longer duration. The shifts may be far more subtle (or completely nonexistent) in people with recently developed type 2 diabetes, which might reduce the voice-recognition technique’s efficacy as a tool for screening or diagnosis. Kaufman told Diabetes Daily that “we have also considered this question and, to address it, we are looking to perform the same study on people with prediabetes.”
With more than 200 million people worldwide unaware that they have type 2 diabetes, there is an immense demand for simpler screening techniques.